ALL-OPTICAL NONLINEAR SNR MONITORING AND LAUNCHED POWER OPTIMIZATION BY ARTIFICIAL NEURAL NETWORKS

In this paper, we present an artificial neural network able to monitor the power ratio between linear noise and nonlinear impairments based on the all-optical power spectral density, i.e., without the need of full coherent reception. The results are used to correct the actual launch power to the optimum one providing a performance gain in terms of signal-to-noise ratio. We validate this monitor experimentally over several configurations varying link lengths and dispersion maps. At the receiver, we use an adaptive carrier phase estimator to partially suppress for nonlinearities, and to introduce a significant alteration of the transmission performance non-trivially related to the all-optical spectrum. The monitor presents the following performance, in terms of power correction prediction: a standard-deviation less than 0.45 dB and a maximum-absolute-error less than 1.65 dB.